Development of the first metabolite-based LC-MSn urine drug screening procedure-exemplified for antidepressants
In contrast to GC-MS libraries, currently available LC-MS libraries for toxicological detection contain besides parent drugs only some main metabolites limiting their applicability for urine screening. Therefore, a metabolite-based LC-MSn screening procedure was developed and exemplified for antidepressants. The library was built up with MS2 and MS3 wideband spectra using an LXQ linear ion trap with electrospray ionization in the positive mode and full-scan information-dependent acquisition. Pure substance spectra were recorded in methanolic solution and metabolite spectra in urine from rats after administration of the corresponding drugs. After identification, the metabolite spectra were added to the library. Various drugs and metabolites could be sufficiently separated. Recovery, process efficiency, matrix effects, and limits of detection for selected drugs were determined using protein precipitation. Automatic data evaluation was performed using ToxID and SmileMS software. The library consists of over 700 parent compounds including 45 antidepressants, over 1,600 metabolites, and artifacts. Protein precipitation led to sufficient results for sample preparation. ToxID and SmileMS were both suitable for target screening with some pros and cons. In our study, only SmileMS was suitable for untargeted screening being not limited to precursor selection. The LC-MSn method was suitable for urine screening as exemplified for antidepressants. It also allowed detecting unknown compounds based on known fragment structures. As ion suppression can never be excluded, it is advantageous to have several targets per drug. Furthermore, the detection of metabolites confirms the body passage. The presented LC-MSn method complements established GC-MS or LC-MS procedures in the authors’ lab.
KeywordsUrine Screening LC-MS Library Metabolite Antidepressants
A major goal in clinical and forensic as well as in doping laboratories is a broad screening procedure for detection of toxic compounds or drugs of abuse. Common GC-MS screening, using comprehensive libraries with electron impact spectra and sophisticated search algorithms [1, 2], provides excellent screening results with the limitation of volatile and apolar compounds. LC-MS overcomes both disadvantages, so that several LC-MS [3, 4, 5, 6, 7, 8] or LC-MSn [9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24] screening procedures using library search were described. Different mass analyzers were used such as triple quadrupoles [10, 16], ion traps [9, 16, 21, 24], hybrids of both techniques [13, 17, 18, 19, 20, 23], or high-resolution time-of-flight analyzers [8, 11, 14, 15]. Procedures based on multiple-reaction monitoring transitions with or without product ion spectra search are focused on a limited number of monitored compounds (target screening). Procedures based on full-scan mode and library search can be used for target as well as for untargeted screening.
In contrast to reference libraries used for comprehensive GC-MS screening procedures , all currently available libraries for LC-MS screening are not focused on metabolite spectra, although they contain some of the main, mostly commercially available, metabolites. Thus, their applicability for urine screening is limited because most toxicologically relevant compounds are excreted into urine more or less exclusively as phase I and/or II metabolites . Detection of metabolites increases the selectivity, allows confirmation of the body passage, and finally, minimizes the risk of false negative LC-MS results possibly caused by ion suppression of the target analyte. Even the risk of false positive results can be reduced considering the metabolite patterns. Therefore, the aim of the presented study was to develop a full-scan, metabolite-based screening procedure focused on urine analysis as a complement to other general unknown LC-UV, GC-MS, and LC-MS screening procedures . This general approach was exemplified for screening for antidepressants because of their high relevance in emergency toxicology.
Chemicals and reagents
The drugs were kindly provided by the corresponding pharmaceutical companies with the following exceptions: Codeine was obtained from Lipomed (Weil am Rhein, Germany), morphine from Promochem (Wesel, Germany), morphine-6-glucuronide from Sigma-Aldrich (Taufkirchen, Germany), and phenylephrine-glucuronide from Toronto Research Chemicals (Toronto, Canada). 4-Hydroxy-3-methoxy-methamphetamine sulfate was synthesized in-house. Ammonium formate (analytical grade) and formic acid (mass spec grade) were obtained from Fluka (Neu-Ulm, Germany), acetonitrile, methanol, and water (all LC-MS grade) from Fisher Scientific (Schwerte, Germany), and all other chemicals (analytical grade) from Merck (Darmstadt, Germany).
All samples were analyzed using a LXQ linear ion trap (ThermoFisher Scientific, TF, Dreieich, Germany) mass spectrometer coupled to a TF Accela ultra HPLC (UHPLC) system consisting of a degasser, a quaternary pump, and an autosampler. Gradient elution was performed on a TF Hypersil GOLD C18 column (100 × 2.1 mm, 1.9 μm). The mobile phase consisted of 10 mM aqueous ammonium formate plus 0.1% formic acid (pH 3.4, eluent A) and acetonitrile plus 0.1% formic acid (eluent B). The flow rate was set to 0.5 mL/min and the gradient was programmed as follows: 0–1.0 min 98% A, 1.0–3.0 min to 90% A, 3.0–5.0 min to 85% A, 5.0–7.5 min to 80% A, 7.5–10.0 min to 75% A, 10.0–11.5 min to 70% A, 11.5–13.0 min to 65% A, 13.0–14.5 min to 50% A, 14.5–16.0 min to 40% A, 16.0–19.0 min to 0% A, 19.0–21.0 hold 0% A. For cleaning the injection system a methanol/water (85:15, V:V) solution was used as follows: needle, three times 3 mL, injection loop, twice 10 μL. Column flushing and re-equilibrating were performed during these cleaning steps. Thus, the total run time per sample was 25 min. The injection volume for all samples was 10 μL each.
The MS was equipped with a TF heated ESI II source, other conditions were as follows: positive ionization mode; sheath gas, nitrogen at flow rate of 34 arbitrary units (AU); auxiliary gas, nitrogen at flow rate of 11 AU; vaporizer temperature, 250 °C; source voltage, 3.00 kV; ion transfer capillary temperature, 300 °C; capillary voltage, 31 V; tube lens voltage, 80 V. Automatic gain control was set to 15,000 ions for full scan and 5,000 ions for MSn. The maximum injection time for full scan (MS1 stage) was set to 100 ms. Collision induced dissociation (CID)-MSn experiments were performed on precursor ions selected from MS1 using information-dependent acquisition (IDA): MS1 was performed in the full-scan mode (m/z 100–800). MS2 and MS3 were performed in the IDA mode: four IDA MS2 experiments (so-called scan filters) were chosen to provide MS2 on the four most intense signals from MS1 and additionally, eight MS3 scan filters were chosen to record MS3 on the most and second most intense signals from the MS2. MS2 spectra were collected with a higher priority than MS3 spectra. Normalized wideband collision energies were 35.0% for MS2 and 40.0% for MS3. Other settings were as follows for MS2: minimum signal threshold, 100 counts; isolation width, 1.5 u; for MS3: minimum signal threshold, 50 counts; isolation width, 2.0 u; for both stages: activation Q, 0.25; activation time, 30 ms; dynamic exclusion mode: repeat counts, 2; repeat duration, 15 s; exclusion list, 50; exclusion duration, 15 s; average full scan to full-scan cycle time, 4 s.
For performance check, a test mixture according to that for GC-MS  (methanolic solution, 1.0 mg/L, of codeine, diazepam, diphenhydramine, haloperidol, methaqualone, morphine, nalorphine, quinine, and strychnine) was injected daily. To ensure standard instrumental conditions for the measurement of product ion spectra, known as the calibration point , the MS2 spectrum of diazepam was compared by library search with that of the library reference spectrum using TF ToxID.
TF Xcalibur 2.1.0 was used for data acquisition, NIST MS Search 2.0 (National Institute of Standards and Technology, Gaithersburg, MD) for library generation, MassFrontier 220.127.116.11 (Highchem, Bratislava, Slovakia) for prediction of MS2 and MS3 fragments, TF ToxID 2.1.1 for target screening in the MS2 screening mode with the following settings: retention time (RT) window, 20 min; RT, 0.1 min; signal threshold, 100 counts; search index, 600; reverse search index, 700, and TF LCquan 2.6.0 software with the following settings: peak detection algorithm, genesis; smoothing, 3; signal-to-noise threshold, 0.5. SmileMS version 1.0.28 (GeneBio, Geneva, Switzerland) was used for target screening using the precursor tolerance option and for untargeted screening without the precursor tolerance option. RT locking was not used. Other settings were as follows: score threshold, 0.1; minimum number peak matches, 0. GraphPad Prism 5.0 (GraphPad Software, San Diego, CA, USA) and QuickCalcs Outlier Calculator (http://www.graphpad.com/quickcalcs/Grubbs1.cfm) were used for statistic calculations.
Mass spectra of the parent compounds were recorded from methanolic stock solutions (1.0 mg/L) after LC separation. The library entries (NIST format) of a compound consist of MS2 or MS3 spectra, names, synonyms, empirical formulas, masses, CAS numbers, and chemical structures. The comment row contains the scan filter information including the M + H mass of the compound. To simplify name searches, the compound name is given again in the synonym field. This contains also names of different compounds (e.g., mitragynine and its 9-O-demethyl metabolite in Ref. ), if they form identical MS3 spectra. These spectra, independent from the MS2 precursor information, lead to the so-called fragment library. Metabolites or artifacts detected in rat or human urine were added to the library. The metabolites were named according to Refs. [1, 29]. If compounds formed the same metabolite(s), their names were added to the synonyms. If the structure of a metabolite or artifact could not clearly be deduced, the corresponding precursor ion was added to the name field.
The interday and intraday reproducibility of the MS2 spectra was tested analyzing a methanolic solution (1.0 mg/L, n = 10) of codeine, diazepam, diphenhydramine, haloperidol, methaqualone, morphine, nalorphine, quinine, and strychnine once a week for a period of 4 weeks. The concentration-dependent reproducibility of the MS2 spectra was tested after 10 injections each of a 0.1, 1.0, and 10 mg/L solution. The criterion of reproducibility was the comparison of the acquired spectra with the corresponding spectra stored in the library expressed as ToxID search index value. The values were processed by GraphPad using one-way ANOVA analysis.
The investigations were performed using urine of male Wistar rats (Ch. River, Sulzfleck, Germany). The different single drugs were administered to different rats in a single 20 mg/kg body mass dose in aqueous suspension by gastric intubation for toxicological diagnostic reasons according to the corresponding German law. Urine was collected separately from the feces over a 24-h period stabilized with sodium fluoride. The samples were directly analyzed or stored at −20 °C until further analysis. Blank rat urine samples were collected before drug administration to check whether they were free of interfering compounds. Additionally, several human urine samples of different patients submitted to the authors’ laboratory for toxicological routine analysis were used.
To 0.1 mL of urine, 0.5 mL of acetonitrile were added. The mixture was shaken on a rotary shaker for 2 min. After centrifugation for 3 min at 10,000×g, 0.5 mL was transferred into a glass vial and evaporated to dryness under a gentle stream of nitrogen at 50 °C. The residue was dissolved in 0.05 mL of a mixture of eluents A and B (1:1; V:V) to ensure sufficient wet ability.
Recovery, matrix effect, process efficiency, limit of detection
Recovery, matrix effect, process efficiency, and limit of detection of the tested compounds (sorted according to RT)
RE (n = 6)
ME (n = 6)
PE (n = 6)
LOD values were determined using urine spiked with spike solution or a 1:10 dilution of that. The urine samples were worked up leading to final urine concentrations of 1.0 or 0.1 mg/L, respectively. All samples were analyzed by the above-described LC-MS method. The criterion for the LOD was defined as that concentration at which in all six samples ToxID still detected the compound.
Results and discussion
The chromatographic system provided sufficient separation and peak shape for drugs of many classes and their metabolites. Wideband activation was used to suppress, e.g. the common loss of water of many compounds in MS2 leading to more fragments in MS2 . The IDA settings allowed triggering even traces of compounds at least in MS2. The dynamic exclusion time was set to 15 s allowing a compound to pass through an MS2MS3MS3 cycle at least two times per peak with a typical width of 30 s. The MS3 spectra were recorded of the most and the second most intense ions of the MS2 spectra improving spectral information. Furthermore, this approach reduced the risk of missing a compound if the underlying MS2 spectrum was overlapped by another compound with the same nominal molecular mass.
A daily performance test for the chromatographic system should be performed providing information on sensitivity, separation power including peak shape. Peak shape may influence IDA MSn scan events so that a compound might be missed. After system maintenance, the MS2 spectrum of diazepam was compared with the library spectrum to check the system integrity according to Hopley et al. . Even using three washing steps, carry-over should be excluded using blank samples to avoid false positive results. Thus, the overall analysis time was about 50 min including the recommended blank sample injection.
Generation of the reference library
MS2 and MS3 spectra of over 700 toxicologically relevant compounds including 45 antidepressants were added to the library after LC separation from methanolic stock solutions (1.0 mg/L). Over 1,600 mass spectra of metabolites or artifacts (about 500 from antidepressants) were recorded from rat or human urine samples after protein precipitation (PP) and LC separation. Only in the early development phase, solid-phase extraction was also used . After careful interpretation of the fragmentation patterns and considering published metabolic patterns [1, 25], these mass spectra were added to the library. In cases of doubt, the interpretation was confirmed by high-resolution mass spectrometry [28, 31, 32] or by GC-MS using the corresponding reference library . In addition, spectra of endogenous biomolecules and common impurities were included for differentiation. So, the current library consists of over 4,300 spectra of over 2,400 different compounds.
A five-mass list of the corresponding MS2 spectra of all studied antidepressants and their phase I and II metabolites is given in the Electronic Supplementary Material Table S1 as part of the reference library in preparation. It contains the compound names, the analyte type, the M + H, and the five most abundant fragment masses, sorted according to ascending mass, and their relative abundances. For reasons of space, synonyms (e.g., mCPP as a trazodone and nefazodone metabolite) are not listed in Electronic Supplementary Material Table S1, but in the online version of the library. As mCPP is a designer drug, but not an antidepressant, it is not listed here, in contrast to synonyms which are antidepressants (e.g., desipramine, also a metabolite of imipramine and lofepramine).
Artifacts formed during the ionization process (e.g. N-oxide dimers) are also listed. Variations of the ionization conditions would reduce the formation of artifacts, but one can imagine that for such a comprehensive screening approach some compromises have to be made. As the artifact formation is not really a problem, as they provide additional targets, the use, e.g., of APCI would lead to other disadvantages (e.g., higher in-source fragmentation with less sensitivity).
The metabolite spectra were recorded from rat or human urine samples following the recommendation of Josephs and Sanders . They state that the best library fits are achieved when a spectrum is acquired under conditions identical to those used for creating the reference spectrum in the library. As described for series of drugs [28, 31, 32, 33, 34, 35], metabolite spectra recorded in rat urine under controlled conditions could also be found in human urine. Possible species differences in glucuronidation versus sulfation [28, 31, 32] can be overcome using the fragment library, because in both cases, the corresponding MS3 spectra are identical caused by neutral loss of the glucuronide or sulfate moiety.
The presented LC-MSn method was not suitable to detect tranylcypromine and its metabolites using wideband MSn spectral information lacking of reproducibility. Other analytical methods, e.g., LC-MS without wideband activation or GC-MS  should be performed alternatively.
Reproducibility of the reference mass spectra
The reproducibility was sufficient for successful library use. However, spectra recorded by different mass analyzer types may vary considerably. Using sophisticated search algorithms such as SmileMS, these variations might be overcome, e.g., by reducing the relative ion abundance information in contrast to the formation of the ions. Similar comparison studies have already been performed [37, 38, 39].
Sample preparation for urine screening
After various preliminary experiments, PP was tested for suitability for non-selective sample workup for comprehensive urine screening providing fast workup advantages, e.g., in emergency toxicology. Hydrolysis of phase II metabolites would increase the amount of the corresponding phase I metabolite and such its detectability. However, the matrix would also increase. If cleavage is necessary, the advantage of LC-MS over GC-MS would be limited.
PP was tested for suitability for non-selective sample workup for comprehensive urine screening detecting a variety of compounds of different drug classes and their metabolites. RE, ME, PE, and LOD of model compounds (drugs and some commercially available phase I and II metabolites) selected according to their chromatographic behavior and chemical structure (different amines, alcohol, phenol, carboxylic acid, N-oxide, conjugates) were determined. Table 1 summarizes the results of the tested parameters.
As given in Table 1, the RE data for PP of about 80% correspond to the percentage of the sample volume taken after centrifugation. The relative high means of the recoveries for some compounds were caused by overlapping matrix peaks with the same M + H full-scan mass in the same two urine samples. The values were tested to be no outliers using Grubbs outlier test. Although at least eight data points per peak were recorded, high coefficients of variation (CVs) were observed, most probably caused by the varying matrix of the tested urine samples of different sources. ME, mainly ion suppression, was observed for several analytes over the whole chromatogram. The PE was overall acceptable for the tested compounds.
The criterion for the LOD was defined as the concentration at which ToxID still detected the compound in all six samples. Common signal-to-noise ratios of the full-scan signals were not applicable for determination of the LODs because the detection of the analyte was not performed based on the analyte’s full-scan mass. The criteria here were the possibility to perform an IDA MSn experiment and the spectra quality which is expressed as ToxID search index value. The spectra quality could often be influenced by coeluting compounds with the same precursor leading to mixed MS2 spectra. Fortunately, several tested metabolites could be detected in most cases. If the detection of one of them is disturbed, another metabolite should be detectable avoiding false negative results. For example, in every of the six spiked urine samples, either venlafaxine or one of its metabolites or artifacts formed during the ionization process (venlafaxine-M/artifact-H2O) could be detected even at the lower LOD of 0.1 mg/L. The same was found for amitriptyline and its metabolites or artifacts.
In summary, PP has proved to be suitable, at least for antidepressants, for routine screening being simple, time- and resources-saving with no limitations concerning pH and extractions properties. As discussed above, the potential limitations of PP described by Li et al.  such as concentrating the interfering compounds increasing the risk of ion suppression or non-triggering IDA MSn experiments are of less importance because the new screening procedure focuses on several targets per drug.
Search algorithms for urine screening
For automatic data evaluation, ToxID and SmileMS were used. ToxID analyses were generated automatically after file recording. The settings were suitable for target screening, based on the NIST search algorithm implemented in the Xcalibur data system. This screening approach allowed detection of the described antidepressants and/or their metabolites in human urine samples. These patient samples, submitted to the authors’ lab, were routinely screened by the GC-MS approach described by Maurer et al. . In addition, several proficiency tests were performed successfully using the new approach. For example, in the GTFCh proficiency testing schemes (http://www.arvecon.com/downloads/pt2010e.pdf) for Qualitative Screening Analysis and Drugs in Urine, antidepressants and in addition, opioids, benzodiazepines, amphetamines could be detected.
Excerpt of a ToxID search result file sorted according to the actual retention time (RT) of an LC-MSn analysis of an authentic patient urine sample after PP presenting the RT, compound names, compound information, m/z of the protonated molecular ion (MS1 full scan), search index (SI), and the reverse search index (RSI) value (max. 999)
MS2_Ampicillin-M (ampicilloic acid)_wideband_35
MS2_Promethazine-M (HO- glucuronide) isomer 1_wideband_35
MS2_Promethazine-M (nor-HO-) isomer 1_wideband_35
MS2_Bromazepam-M (HO- glucuronide) isomer 1_wideband_35
MS2_Paroxetine-M (demethylenyl-methyl- glucuronide) isomer 2_wideband_35
MS2_Promethazine-M (HO-) isomer 2_wideband_35
MS2_Trimipramine-M (bis-nor-HO- glucuronide)_wideband_35
MS2_Trimipramine-M (HO- glucuronide) isomer 1_wideband_35
MS2_Trimipramine-M (nor-HO- glucuronide)_wideband_35
MS2_Chlorprothixene-M (HO- sulfate) isomer 1_wideband_35
MS2_Endogenous biomolecule (286)_wideband_35
MS2_Endogenous biomolecule (464)_wideband_35
MS2_Paroxetine-M (demethylenyl-methyl-) isomer 2_wideband_35
MS2_Diazepam-M (HO-methyl- glucuronide)_wideband_35
MS2_Levomepromazine-M (nor-O-demethyl- glucuronide)_wideband_35
MS2_Mirtazapine-M (HO-ring- glucuronide)_wideband_35
MS2_Tolazamide-M (HO-aryl- glucuronide)_wideband_35
MS2_Endogenous biomolecule (358) isomer 1_wideband_35
MS2_Endogenous biomolecule (358) isomer 2_wideband_35
MS2_Trimipramine-M/artifact (N-oxide-) dimer_wideband_35
SmileMS  in the target screening mode showed similar results as ToxID. All compounds detected by ToxID could also be detected using SmileMS with the exception of lidocaine whose MS2 spectrum contains only one predominant ion (m/z 86) and attributed therefore to a poor score by SmileMS.
ToxID and SmileMS were both suitable for target screening with some pros and cons. A systematic comparison study is in progress. However, as ToxID evaluated only MS2 data, search results particularly of phase II metabolites must be confirmed by manual MS3 data evaluation using Xcalibur. Otherwise, there is a risk of false positive results because after neutral loss of the conjugates, the MS2 often contained only a few ions. For example, bromazepam-M (HO-glucuronide) isomer 1, chlorprothixene-M (HO- sulfate) isomer 1, and tolazamide-M (HO-aryl-glucuronide) listed in Table 2, could not be confirmed manually by a skilled user. The corresponding drugs or their metabolites could also not be detected by GC-MS.
However, SmileMS used without any precursor selection was able to perform untargeted screening matching all spectra of the corresponding run with all library spectra. Thus, SmileMS was able to detect also unknown phase II metabolites by comparison their MS3 spectra with the MS2 spectra of underlying known phase I metabolites. In addition, using the fragment library, SmileMS was also able to detect automatically unknown compounds with certain (group-specific) structures. This unknown screening feature was used successfully in current metabolism studies as well as in routine screening of patient samples. For example, a SmileMS search result indicating unknown indeloxacine metabolites in rat urine is depicted in Electronic Supplementary Material Fig. S1. SmileMS search score, experimental and reference data for RT and m/z as well as their differences (delta) are given. The marked line indicates an oxidized metabolite (+16 u) not yet stored in the library. Nevertheless, SmileMS was able detecting this unknown indeloxacine metabolite using the fragment library. The structures of such unknown compounds were deduced as described above based on detailed interpretation of the MSn spectra. Numerous metabolite spectra were identified in routine urine samples using SmileMS in the untargeted screening mode and finally added to the library.
In our study, only SmileMS was suitable for untargeted screening, because it is not limited to precursor selection. ToxID has no option for unknown screening procedures, because in ToxID, only the spectra saved in the corresponding CSV file were analyzed. According to the MS2-independent MS3 spectra implementation in the NIST library, ToxID analysis was performed in MS2 confirmation mode only. To use even MS3 confirmation, a complete implementation of all MS3 spectra with the correct MS2 precursors had to be implemented in the ToxID CSV file. Nevertheless, this procedure would also lead only to a target analysis.
However, SmileMS was able to detect unknown compounds with certain (group-specific) structures stored in the fragment library. In addition, SmileMS allowed detection of the target analyte in mixed MS2 spectra using the MS3 spectra ion if one of the most and/or second most abundant ions was formed from the corresponding target analyte. In general, all automatically generated screening results must be evaluated by a skilled user , because limitations of the search algorithm and the library can never exclude false results.
Detection of metabolites allows confirming of the body passage, monitoring an intake of compounds which are excreted completely metabolized, and reducing the risk of false negative LC-MS results possibly caused by ion suppression of the target analyte. Because ion suppression can never be completely excluded, the best possibility to overcome this problem is to have more targets per drug to screen for. Even, the risk of false positive results can be reduced considering the metabolite patterns. For example, if mCPP is detected in urine, it has to be differentiated whether the designer drug mCPP was taken or the antidepressants trazodone or nefazodone. This can easily be differentiated by detection of other unique trazodone or nefazodone metabolites. Therefore, the detection of metabolites may increase the selectivity of the screening procedure. Finally, all compounds present in urine have to be identified to be sure that these are no (further) compounds relevant in clinical and forensic toxicology and doping control.
The developed metabolite-based LC-MSn screening method was suitable for urine screening as described and exemplified here for antidepressant drugs. In contrast to current LC-MSn libraries, which contain only some of the main, mostly commercially available, metabolites, the presented library contains all phase I and phase II metabolites of the studied drugs detectable by this approach and is, therefore, more suitable for urine screening. Additionally, the presented screening concept does allow detecting automatically unknown compounds/metabolites based on known fragment structures. The presented LC-MSn method was suitable to complement established GC-MS or LC-MS screening procedures used in the authors’ lab.
The authors thank Benjamin Honold, Golo M. Meyer, Peter Niklas, Frank T. Peters, Anika-Anina Philipp, Andrea E. Schwaninger, Carsten Schröder, Gabi Ulrich and Lydia Zweigler as well as Kornelia Weidemann and Edeltraud Thiry (ThermoFisher Scientific Instruments, Dreieich, Germany), and Pierre-Alain Binz, Nicolas Budin, Yann Mauron, and Roman Mylonas (GeneBio, Geneva, Switzerland) for their support.
- 1.Maurer HH, Pfleger K, Weber AA (2007) Mass spectral and GC data of drugs, poisons, pesticides, pollutants and their metabolites. Wiley-VCH, WeinheimGoogle Scholar
- 3.Rittner M, Pragst F, Bork WR, Neumann J (2001) J Anal Toxicol 25:115–124Google Scholar
- 5.Venisse N, Marquet P, Duchoslav E, Dupuy JL, Lachatre G (2003) J Anal Toxicol 27:7–14Google Scholar
- 9.Fitzgerald RL, Rivera JD, Herold DA (1999) Clin Chem 45:1224–1234Google Scholar
- 25.Baselt RC (2009) Disposition of toxic drugs and chemicals in man. Biomedical Publications, Foster CityGoogle Scholar
- 29.Maurer HH, Pfleger K, Weber AA (2007) Mass spectral library of drugs, poisons, pesticides, pollutants and their metabolites. Wiley-VCH, WeinheimGoogle Scholar
- 32.Philipp AA, Wissenbach DK, Weber AA, Zapp J, Maurer HH (2010) J Mass Spectrom, in pressGoogle Scholar
- 34.Maurer HH, Bickeboeller-Friedrich J (2000) J Anal Toxicol 24:340–347Google Scholar
- 36.Schwaninger AE, Meyer MR, Maurer HH (2010) J Anal Toxicol 34:45–48Google Scholar